By Deborah Borfitz
October 15, 2019 | The U.S. Food and Drug Administration (FDA) has been unusually busy over the past two years on programs and projects advancing regulatory science around in vitro diagnostics (IVD), including digital pathology. It parallels an unprecedented pace of product-related developments, including several firsts—approval of a fully implantable glucose monitoring device and direct-to-consumer test for detecting genetic variants, recognition of a genetic variant database and release of a white paper on artificial intelligence (AI).
Two FDA representatives highlighted recent activities and agency perspectives during a pair of sessions at the 2019 Next Generation Dx Summit in Washington, D.C.—Katherine Donigan, acting director of personalized medicine in the Office of Health Technology (OHT7, formerly the Office of In Vitro Diagnostics and Radiological Health), Center for Devices and Radiological Health (CDRH), and Shyam Kalavar, senior scientific reviewer in the molecular pathology and cytology branch, in the OHT7's division of molecular genetics and pathology.
During her plenary keynote presentation, Donigan focused on the outreach, policy, and research activities of OHT7 and made note of the FDA's new "total product life cycle" approach that integrates premarket and post-market data about medical devices. The approval of 106 novel devices in 2018 made it a record year for device innovation, she says, surpassing the 40-year record set by the FDA in 2017.
Guidance and Strategy
Other 2018 achievements were completion of two guidance documents on next generation sequencing (NGS)-based tests, one of which describes what the FDA looks for in premarket submissions to determine a test's analytical validity, Donigan says. It also issued draft guidance on devices to be included in its 510(k) Third Party Review Program.
The 510(k) Third Party Review Program is an alternative review process for manufacturers of certain low-to-moderate risk medical devices, she says, and is designed to yield more rapid 510(k) decisions and allow the FDA to focus its resources on higher risk devices. Submissions are made to accredited third party review organizations rather than directly to the FDA.
One of the latest to be accredited is the New York State Department of Health, whose expertise includes devices that do NGS-based tumor profiling tests, Donigan says. Program challenges have been the quality of submissions that often necessitates re-review and limited public information to support the link between genetic variation and health.
Last November, the agency introduced draft guidance on using the Dual 510(k) and CLIA Waiver by Application pathway. The agency also recently finalized guidance around its Breakthrough Device Designation, she says. The two main criteria for inclusion are that the device represents breakthrough technology with no approved or cleared alternatives and its availability is in the best interests of patients.
As of late August, 156 devices have been so designated, Donigan adds, 50 of which are diagnostics. Eleven have received market authorization and six are in the premarket approval process. Among the early successes were FoundationOne CDx, the first FDA-approved broad companion diagnostic (CDx) for solid tumors, and Banyan Brain Trauma Indicator, the first blood test to evaluate concussion.
FDA actively encourages private- and public-sector members to work on medical device challenges together via collaborative communities, which may or may not include the FDA and "could exist indefinitely," says Donigan. CDRH would like to see 10 collaborative communities formed by the end of 2020 and has produced a toolkit to facilitate their development and success.
To date, the FDA—long an advocate of precision medicine—has approved 37 unique IVD CDx and over 100 human nucleic acid-based tests, says Donigan. Thirty cancer therapeutic products now have a companion diagnostic.
As summarized in a fact sheet focused on three key submissions that have been cleared for marketing (Thermo Fisher Scientific's Oncomine Dx Target Test, MSK-IMPACT and Foundation Medicine's FoundationOne CDx), the FDA takes a tiered approach to regulating tumor profiling NGS tests with differing levels of required evidence, says Donigan.
The agency's vision for regulation of NGS-based IVD for germline diseases, Donigan adds, are technical and analytical standards that device developers could meet in lieu of making a premarket submission.
IVD sponsors using genetic variant databases to support clinical validity are advised to use recognized, publicly accessible databases, she says.
The FDA has a voluntary pathway for databases to receive recognition and ClinGen, founded by the National Human Genome Research Institute, was the first to be approved on Dec. 4, 2018. "We're fully transparent about how we do the review and who does the curation and we do spot checks," says Donigan. ClinGen is being used by IVD developers to assure data reliability "in certain contexts… so they don't have to generate the same data over and over on their own."
Analytical guidance on NGS-based IVD only applies to targeted or whole exome sequencing tests, she continues, and makes recommendations about how devices can be designed, validated, and labeled as well as meet the standards for possible 510(k) exemption.
The cloud-based precisionFDA provides a collaborative omics platform to analyze, integrate, and compare datasets, says Donigan. The agency has launched several challenges focused on the reliability and accuracy of NGS bioinformatics pipelines to promote the platform, the end goal being to "drive regulatory science in this space."
The OHT7 is actively involved in the development of NGS reference sample sets that can be used to create an "evolving truth sequence" for validating NGS-based oncology tests, Donigan says, and they'll be accessible on the precisionFDA platform. The first phase of the project involves a prioritized set of samples and variants, and phase two will look at products and characterization.
Five community-driven reference materials have been developed to date and other surrogate reference sample projects are underway, she says, including the Genome in a Bottle consortium hosted by the National Institute of Standards and Technology and supported by CDRH.
The public-private Medical Device Innovation Consortium (MDIC) is a key collaborator, says Donigan, and is championing the advancement of regulatory science through its Clinical Diagnostics program. Focus areas include projects using surrogate samples to support product development and defining analytical validity study designs for fingerstick point-of-care devices. Meanwhile, the FDA's SHIELD (Systemic Harmonization & Interoperability Enhancement for Lab Data) initiative is supporting efforts to leverage non-traditional IVD data sources to support decision-making throughout the total product life cycle.
WSI Validation Studies
Regulation of digital pathology for primary diagnosis is the specialty of Kalavar, who notes that most such systems would be considered class II medical devices cleared through the 510(k) pathway. The first to market in 2017, via the de novo premarket review pathway, was the Philips IntelliSite Pathology Solution (PIPS). It served as the predicate device for the second-to-market device, the Leica Biosystems Aperio AT2 DX System, approved via a 510(k) substantial equivalence determination earlier this year, he says.
The PIPS allows pathologists to view and evaluate digital images of tissue slides that would otherwise be appropriate for manual visualization by conventional brightfield (light) microscopy, Kalavar says. The whole slide imaging (WSI) system consists of an ultra-fast scanner, an image management system and a display.
The FDA likes to see four types of validation studies—technical performance assessment (TPA), analytical, clinical and human factors evaluation, he continues. Guidance issued in 2016 covers component- and system-level testing recommendations for TPA. Human factors assessment studies for WSI should evaluate user-related hazards and the ability to perform essential tasks.
The analytical study on PIPS submitted by Philips looks at 21 histopathologic features, seven per magnification and each selecting three different organs, says Kalavar. Each organ has six fields of view, one per slide, involving many reading pathologists.
Clinical studies on WSI devices should mimic the intended setting, demonstrate non-inferiority to using an optical microscope and evaluate differences in discordance rates between manual digital and manual optical when compared to reference, Kalavar says. Philips accomplished this for the PIPS with a four-site study involving 16 pathologists and 2,000 archived cases representing all major organs, a washout period of four weeks and pathologists doing adjudication of minor and major discordance. The system is designed to arrive at "ground truth" via a consensus diagnosis.
Modifications to any FDA-cleared diagnostic WSI system, such as a change in display, would require a 510(k) application, Kalavar says. The agency's thinking regarding upgrades to the image management system is that there be an agreement in place between the WSI and software manufacturer about who will take responsibility for any changes to the scanner and image format.
To promote the development and availability of safe and effective interoperable medical devices, the FDA issued guidance in 2017 that includes design considerations and recommendations for contents of pre-market submissions, Kalavar notes. FDA-recognized standards are encouraged and, to that end, Digital Imaging and Communications in Medicine (DICOM) has established a working group to expand its standard to pathology imaging.
The 510(k) regulatory pathway requires that an IVD-labeled WIS system serve as the predicate device, says Kalavar. In some cases, a clinical study may be needed. A cleared WSI system that later has AI software added to it constitutes a new device, he adds.
As indicated in the recently published white paper, the FDA is taking a risk-based approach to AI-based software as a medical device. Digital pathology systems using AI will likely be evaluated as a class II device via the de novo pathway, Kalavar says. However, if an adaptive algorithm is involved that continues learning after a device is cleared for marketing, the device could be kicked into a higher risk category—especially as it seeks to replace the standard of care and take pathologists out of the equation. "I'm not sure we're ready for that."
Correction: An earlier version of this story referred to an interview with Katherine Donigan and Shyam Kalavar. Instead this story reflects the presentation the two gave at the event.